5 Signs You’re Automating the Wrong Process
Why AI doesn't fix broken processes, and what leaders should address first.
There's a sentence I keep hearing as organizations move faster into AI and automation: "Can we automate this?"
Sometimes the answer is absolutely yes, but I think we're asking that question too early. I'm excited about what AI can do. I also don't believe in using technology just because we can.
After years of leading transformation across HR, IT, finance, and shared services, I've learned that technology rarely fixes a process problem by itself. In fact, one of the fastest ways to make a bad process worse is to automate it.
Now the confusion moves faster, the unnecessary approvals happen faster, and the exceptions multiply faster. And the frustrating employee experience becomes more efficient at frustrating people.
So before I ask whether we can automate something, I want to understand something much more basic: Should this process work this way at all?
Here are five signs the answer might be no.
1. No One Can Explain Why the Process Exists
This is one of my favorites. Ask why a step exists and someone says: "That's just how we've always done it." Or: "I think Legal wanted that." Or the classic: "The system requires it." Maybe it does. But maybe it doesn't.
I've been part of enough process reviews to know how easily requirements accumulate over time. A control gets added because of something that happened five years ago. A leader asks for another approval. Someone creates a spreadsheet to compensate for a system limitation. Then another team builds a workaround around the spreadsheet. Eventually, no one remembers why half of it exists.
Then someone suggests AI.
That should be a warning sign. If you cannot explain the purpose of each major step, do not automate it yet. First ask: What risk is this step managing? What decision is being made? Who actually needs this information? What would happen if we removed it?
Sometimes the best automation opportunity is deletion.
2. Your Process Depends on Heroics
You probably know this process.
It "works" because Susan knows exactly who to call. Or because someone in payroll keeps a personal spreadsheet. Or because an HR team member remembers that one business unit always submits something differently.
On paper, the process looks fine. In reality, people are holding it together.
I've seen this repeatedly in shared services environments. The documented process tells you what is supposed to happen. The people doing the work tell you what actually happens. That gap matters.
If your process requires institutional memory, side conversations, manual intervention, or one experienced employee who "just knows," automating the documented workflow may actually remove the very things keeping it functional. Before you automate, sit with the people doing the work. Ask them where they intervene. Ask what they fix manually. Ask which cases never follow the documented path. Those conversations will tell you more about whether a process is ready for automation than the process map alone.
3. You Have More Exceptions Than Rules
There will always be exceptions. The problem is when the exceptions have quietly become the process.
I've seen organizations build increasingly complicated workflows because every unusual situation gets its own branch.
If employee type A, send here. Unless they're in this location. Unless they report to this leader. Unless compensation exceeds this threshold. Unless it's a transfer. Unless...
At some point, you are not automating a process. You are encoding years of unresolved decisions. AI can be very good at handling complexity, but that doesn't mean every complexity deserves to survive. Before building automation around dozens of exceptions, look at the exceptions as data. Why do they exist? Which ones happen frequently enough that they should become part of the standard process? Which ones represent legitimate business needs? And which ones exist because nobody has wanted to make a decision?
If we haven't made the decisions yet, automation isn't going to make them for us.
4. You're Solving for HR Instead of the Person Using the Process
This one can be uncomfortable. A process can be incredibly efficient for HR and still be terrible for employees and managers. We centralize intake, create forms, build workflows, add required field, introduce portals, and route everything correctly. And technically, the process works.
But then a manager needs three systems and 14 clicks to make a basic employee change. Or an employee has no idea whether their request disappeared into a queue. That is not transformation. It is a well-organized internal process that forgot who it serves.
Before automating, I like to look at the process from both directions. What does HR need to complete the work correctly? And what does the employee or manager need to get through it easily and understand what happens next? Those are not always the same thing.
Good automation should reduce effort on both sides, not simply transfer administrative work from HR to employees and managers.
5. You Can't Define What "Better" Means
This may be the biggest one.
"We want to use AI." Great. For what?
"We want more automation." Why?
If the answer is simply efficiency, I would keep asking questions. Are we trying to reduce cycle time? Improve first-time accuracy? Reduce rework? Create a better employee experience? Free HR capacity for higher-value work? Improve compliance? Give managers faster answers? Those are very different problems, and they may require very different solutions.
One lesson I've learned from transformation work is that technology projects get into trouble when the technology becomes the outcome. AI is not the outcome. Automation is not the outcome. The outcome is the business or human problem you are trying to solve. If you cannot define what should be better when you are finished, you probably are not ready to automate.
What I'd Do Before Automating Anything
I've become much more disciplined about this over time. Before looking at tools, I would ask five questions:
1. What problem are we actually solving?
2. Which steps could disappear entirely?
3. Where are people compensating for a broken process today?
4. Which exceptions are legitimate, and which are historical baggage?
5. How will we know the experience actually improved?
Only then would I start asking what technology can do. And I think we get ourselves into trouble when we do those things in the opposite order.
Challenging a Bit of Conventional Wisdom
There is enormous pressure right now to show that your organization is "doing AI."
I understand it. Boards are asking. CEOs are asking. Employees are experimenting with tools on their own. Vendors are promising major productivity gains. And there are some incredible opportunities.
I'm not suggesting organizations should sit on the sidelines. But speed of adoption is not the same as maturity.
Being thoughtful about where you use AI is not the same as being slow to adopt it. Sometimes the smarter decision is to not automate something yet.
Fix the ownership. Remove unnecessary steps. Clarify decision rights. Standardize what should be standardized. Then automate.
I want organizations using AI. I just want us using it on the right problems. Because AI can make a good process dramatically better. But it can also make a bad process dramatically faster.
And faster isn't always progress.
#HRRealTalk
About the Author
Amy Wang shares real-world insights of organizational transformation across HR, IT, finance, and shared services. With experience spanning higher education, healthcare, and automotive industries, she brings a grounded perspective to leading change in complex environments. Amy also serves as a strategic advisor on AI integration, helping organizations align technology with workforce strategy. She started using #HRRealTalk to open up more honest conversations about leadership, change, and the human side of complex systems. She writes about what actually works, lessons learned, and how to lead with both clarity and empathy.
Connect with Amy on LinkedIn: linkedin.com/in/amywang168
Amy Wang | Voice of Real Talk in HR and Organizational Transformation
